Model Summary Bar Chart
📊 What is this?
Model Summary tells you the model's indicators in text and numbers; the Model Summary Bar Chart takes the two most critical of them—— R² and Q²——and plots them as side-by-side bars, so that you can see the gap at a glance.
It is best at exposing one problem: overfitting.
🧐 How to read?
The chart shows a bar comparison of R² and Q²:
| Indicator | Meaning | Expectation |
|---|---|---|
| R² | The model's ability to explain the training data | The higher the better |
| Q² | The model's ability to predict unseen data | The higher the better |
The core reading is to look at "how much the two bars differ":
| Pattern | Diagnosis |
|---|---|
| R² and Q² keep pace with each other | ✅ Healthy model, fitting ability matches generalization ability |
| R² is clearly higher than Q² | ⚠️ Overfitting——the model has memorized the noise as well |
| Both bars are very short | ⚠️ Underfitting——the wrong variables were chosen, or the relationship is inherently weak |
| Q² is negative or close to 0 | ❌ The model has no predictive value |
💡 Rule of thumb:
R² - Q² < 0.2is usually acceptable; once the gap exceeds 0.3, you need to be on guard.
🛠️ How to use?
- As the first health check of the model: after Fit is complete, drag this chart out first and glance at the bar heights
- Verify the cleaning effect: look once before and after deleting data——if the Q² bar has clearly grown taller, the points you removed were real noise
- Help decide the number of components: observe repeatedly together with C+1 / C-1 to find the component count where the gap between R² and Q² is smallest
Switching indicator type
The chart supports choosing the indicator type to display: R² only, Q² only, or a comparison of both.
🎯 Division of labor with "Model Summary"
| Chart | Presentation | Suitable scenario |
|---|---|---|
| Model Summary | Complete indicator list | When you need to see all indicators, the optimal component count, and CV details |
| Model Summary Bar Chart | R² / Q² bar comparison | Quickly judge whether overfitting occurs, good for putting into reports |
💡 Use the bar chart for reporting, use the summary table for troubleshooting——the bar chart has stronger visual impact, and non-technical colleagues can also see at a glance that "the two bars differ by this much, so there is a problem".